Repo used the DataDays 2021 talk -> Genetic Programming for Feature Engineering: An application in Trading System with Cryptocurrencies
Non-gradient-based optimization method, useful when having cost functions that are potentially non-diferentiable, discontinuous, with "cliffs" or "steps".
To evolve the topology of a neural network in order to explore more variations of architectures. Insted of minization of a cost function, the goal is to maximize a fitness function.